The tech industry is rife with examples of manufacturers cannibalizing profitable but aging technology in favor of new and improved systems. And while it's too early to say whether Fibre Channel over Ethernet (FCoE) will put an end to native Fibre Channel hardware in the data center, there are a number of compelling reasons to think it could happen relatively quickly.
The heart of the matter is whether data center operators will still find value in discrete FC networks once they have the option of layering it onto the Ethernet. So while most pundits are busy trying to figure out whether FCoE is better or worse than iSCSI, I think an equally intriguing question is whether FCoE is better or worse than native FC.
Ultimately, that question won't be answered until the FCoE standard is ratified and systems are being deployed in real-world environments. But it is possible to piece together some strong indications based on what we know of the standard so far and the initial devices that have made it off the production line.
Jay Kidd, chief marketing officer at NetApp, lays out the primary benefits of FCoE in this article on eWeek. He's looking at major improvements in data center efficiency through the implementation of lossless 10 GbE fabrics, transparent storage management, lower capital and energy costs and increased application availability. And with new Ethernet extensions like Data Center Bridging (DCB) and Converged Enhanced Ethernet (CEE), FCoE will have access to logical pause capabilities and other traffic prioritization techniques.
In a number of ways, combining Fibre Channel and the Ethernet overcomes some of the deficiencies of both technologies, according to Joseph F. Kovar over at ChannelWeb. Native Fibre Channel performance lags 10 GbE in many respects, particularly as the industry preps for 40 and even 100 GbE. Fibre Channel also is limited to about six miles, after which you need an IP gateway to communicate with remote SANs.
Of course, if the idea of a unified fabric appeals to you, then Voltaire CEO Ronnie Kenneth wants you to consider both FC and Ethernet sitting on top of a 20 Gbps Infiniband network. He says the technology is ready to go today-no need to wait for standards committees to dot every i and cross every t. The company just came out with new lines of 10 GbE and FC storage gateway systems aimed at linking corporate LANs through the company's director-class Infiniband switches.
Regardless of the risks to established product lines, virtually everyone with a stake in Fibre Channel is moving headlong into FCoE. Earlier this month, I asked Shaun Walsh, vice president of corporate marketing at Emulex, whether native Fibre Channel was in danger of fading away. This is what he said:
"The world doesn't know yet. FCoE is new technology and we are a strong advocate for it, but the market place will ultimately decide. But you have to remember that this is not a NIC technology, it's an HBA technology. FCoE will not be on every port in the data center. It will still be a premium service. The difference between a NIC and an HBA is how it deals with error handling, software management and storage management. The values of those things don't change whether you're on the Ethernet or native. We don't expect the market to change radically, so there's no reason to expect the current investment in Fibre Channel equipment won't be valuable in the future because FCoE will be fully compatible with 8G and 16G Fibre Channel."
In the short term, I can easily imagine mixed FC/Ethernet environments. But in the longer term, I still come back to the question, why would anyone want to invest in more Fibre Channel hardware when they can have all that and more on the Ethernet?